International Journal of Scientific Research and Engineering Development

The accurate valuation of residential property is a cornerstone of India's rapidly growing real estate market, a significant factor in its economic stability. Traditional appraisal methods, often reliant on subjective expert judgment, struggle to capture the complex, non-linear dynamics of diverse and fastpaced metropolitan housing markets. This paper presents a comparative analysis of machine learning models for house price prediction, leveraging a comprehensive dataset of properties from four major Indian metropolitan cities: Mumbai, Delhi, Chennai, and Kolkata. The study implements and evaluates three distinct algorithmic approaches: Multiple Linear Regression, serving as a traditional baseline; Random Forest, a robust ensemble bagging technique; and Gradient Boosting, a state-of-the-art algorithm implemented using Python's Scikit-learn and TensorFlow libraries. A rigorous data preprocessing pipeline was developed to handle missing values, encode heterogeneous features, and normalize the data distribution. The models were evaluated using standard regression metrics, including R-squared, Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). The experimental results demonstrate the significant superiority of ensemble methods over the linear model. The Gradient Boosting model achieved the highest performance, yielding a strong R-squared value and the lowest RMSE, indicating its powerful predictive capability. Feature importance analysis from the best-perfoming model revealed that the property's location (city), total area, and the number of bedrooms (BHK) are the most influential determinants of price. This study confirms that advanced machine learning techniques provide a more accurate and robust framework for automated property valuation, offering valuable insights for participants in the Indian real estate market.